Extracting outliers from linear regression analysis

[kourosh kalayehParticipant]

Hi,

I was wondering how I can extract the outliers from the fit command so I can show them on the graph.  Below is the example of what I am trying to do,

In this screenshot I have fit command and I have select to remove the outliers from the fit,

Now, how can I extract these outliers (or their row numbers, etc.) so I can get something like below (use x to identify the outliers),

 

Thanks as always,
Kourosh

[dgteamModerator]

You can use the Extract as Column option to get a column that shows the values included in the fit.  Rows of data that are not included will be blank in the extract columns.

Then you could layer two points commands to indicate the data in the fit and the outliers.

 

Does this help?

 

[kourosh kalayehParticipant]

Hi,

Thanks for the response.

It sort of works, but if I understand your method correctly, the two points command need to have same marker so one can go on top of the other.  Ideally, I want to have something similar to my screenshot where I can cross the outliers explicitly.

Thanks,
Kourosh

[dgteamModerator]

I see what you mean.  Try this …

Use a logical statement in an Expression column to check if “x” and “X Values” are the same.  The statement will return “1” when they match and “0” when they don’t.

Then you have a column of ones and zeros that identifies the outliers.  Use this “Included” column in a mask in a separate points command and set the marker style  to a symbol to mark them as crossed out.

It seems like having an “Include” column as one of the “Extract column” options would be useful for future versions.

[kourosh kalayehParticipant]

Great, perfect work around.

I could not agree more on having “Include” option for “Extract column” in future versions.

Thanks for making DataGraph the best graphing software.

Kourosh

[Author]

Posts